Hospitalization scenarios

Scenario definitions

• Tree canopy

  • Population-based Scenario: AI: Increase by 10% in all zip codes

  • Targeted

    • Scenario AII1: Increase by 10% in zip codes in the lowest 1/5th of current TC cover (i.e. <=20th pctile)

    • Scenario AII2: Increase by 10% in zip codes in the highest 1/5th of the Social Vulnerability Index (i.e. >80th pctile)

    • Scenario AII3: Increase by 10% in zip codes in the highest 1/5th of hospitalization burden (i.e. >80th pctile)

  • Proportionate-universalism

    • Scenario AIII1: Increase by 10% for bottom 1/5th of current TC cover… down to 2% for top 1/5th

    • Scenario AIII2: Increase by 10% for top 1/5th of SVI … down to 2% for bottom 1/5th

    • Scenario AIII3: Increase by 10% for top 1/5th of hospitalization burden … down to 2% for bottom 1/5th

• Impervious surface cover

  • Population-based: Scenario BI: Decrease by 10% in all zip codes

  • Targeted

    • Scenario BII1: Decrease by 10% in zip codes in the highest 1/5th of current imperv cover (i.e. >80th pctile)

    • Scenario BII2: Decrease by 10% in zip codes in the highest 1/5th of the Social Vulnerability Index (i.e. >80th pctile)

    • Scenario BII3: Decrease by 10% in zip codes in the highest 1/5th of hospitalization burden (i.e. >80th pctile)

  • Proportionate-universalism

    • Scenario BIII1: Decrease by 10% for top 1/5th of current imperv cover … down to 2% for bottom 1/5th

    • Scenario BIII2: Decrease by 10% for top 1/5th of SVI … down to 2% for bottom 1/5th

    • Scenario BIII3: Decrease by 10% for top 1/5th of hospitalization burden … down to 2% for bottom 1/5th

Plots

Histograms of IRR, IRD, and perc. diff

Facet by type of intervention (impervious surfaces vs tree canopy) and by scenario type - Population-based, Proportionate Universalism, Targeted

facet_histogram_fun=function(df){
  df %>% 
      ggplot(aes(value))+
      geom_histogram()+
      facet_grid(
        rows=vars(scenario_type_7_abbrev),
        cols=vars(scenario_intervention)
      )
}
hosp_all_long %>% 
  filter(measure=="irr") %>%
  facet_histogram_fun()+
  xlab("IRR")

#All IRD
hosp_all_long %>% 
  filter(measure=="ird") %>%
  facet_histogram_fun()+
  xlab("IRD")

#Exclude outliers
hosp_all_long %>% 
  filter(measure=="ird") %>%
  filter(value<0.005) %>% 
  facet_histogram_fun()+
  xlab("IRD")

hosp_all_long %>% 
  filter(measure=="pd") %>%
  facet_histogram_fun()+
  xlab("PD")

Bar charts

Note I took the mean of each measure over all zip codes. For the ratios and percent differences, I would suggest calculating the mean a different way: calculate the total number of baseline hospitalizations and the total number of hospitalizations prewvented - over all zip codes and then calculate the corresponding measure (IRR or PD).

I think the means would be the same for the IRDs but not for the IRRs and PDs

Mean IRR

hosp_all_long %>% 
  filter(measure=="irr") %>% 
  group_by(measure, scenario) %>% 
  summarise(irr_mean=mean(value,na.rm=TRUE)) %>% 
  ungroup() %>% 
  left_join(lookup_scenario, by ="scenario") %>% 
  ggplot(aes(y=irr_mean,x=scenario_type_7_abbrev))+
  geom_col()+
  labs(x="Scenario", y="Mean IRR over ZCTA")+
  facet_grid(
    #Move facet down to x-axis
    #https://stackoverflow.com/questions/67519146/bar-plot-with-named-groups-on-x-axis-in-ggplot2
    cols=vars(scenario_intervention),
    scales="free_x",
    space="free_x",
    switch="x"
  )+
  theme(panel.spacing = unit(0, units = "cm"), # removes space between panels
        strip.placement = "outside", # moves the states down
        strip.background = element_rect(fill = "white") 
  )

## `summarise()` has grouped output by 'measure'. You can override using the
## `.groups` argument.

Mean IRD

hosp_all_long %>% 
  filter(measure=="ird") %>% 
  group_by(measure, scenario) %>% 
  summarise(ird_mean=mean(value,na.rm=TRUE)) %>% 
  ungroup() %>% 
  left_join(lookup_scenario, by ="scenario") %>% 
  ggplot(aes(y=ird_mean,x=scenario_type_7_abbrev))+
  geom_col()+
  labs(x="Scenario", y="Mean IRD over ZCTA")+
  facet_grid(
    #Move facet down to x-axis
    #https://stackoverflow.com/questions/67519146/bar-plot-with-named-groups-on-x-axis-in-ggplot2
    cols=vars(scenario_intervention),
    scales="free_x",
    space="free_x",
    switch="x"
  )+
  theme(panel.spacing = unit(0, units = "cm"), # removes space between panels
        strip.placement = "outside", # moves the states down
        strip.background = element_rect(fill = "white") 
  )

## `summarise()` has grouped output by 'measure'. You can override using the
## `.groups` argument.

Mean PD

hosp_all_long %>% 
  filter(measure=="pd") %>% 
  group_by(measure, scenario) %>% 
  summarise(pd_mean=mean(value,na.rm=TRUE)) %>% 
  ungroup() %>% 
  left_join(lookup_scenario, by ="scenario") %>% 
  ggplot(aes(y=pd_mean,x=scenario_type_7_abbrev))+
  geom_col()+
  labs(x="Scenario", y="Mean PD over ZCTA")+
  facet_grid(
    #Move facet down to x-axis
    #https://stackoverflow.com/questions/67519146/bar-plot-with-named-groups-on-x-axis-in-ggplot2
    cols=vars(scenario_intervention),
    scales="free_x",
    space="free_x",
    switch="x"
  )+
  theme(panel.spacing = unit(0, units = "cm"), # removes space between panels
        strip.placement = "outside", # moves the states down
        strip.background = element_rect(fill = "white") 
  )

## `summarise()` has grouped output by 'measure'. You can override using the
## `.groups` argument.

Static maps

Ideas for static maps of measures using facet plots

IRR

Population-based scenarios: Impervious surfaces vs. Tree canopy

Proportionate Universalism and Targeted Scenarios

Impervious Surfaces

Tree canopy

IRD

Population-based scenarios: Impervious surfaces vs. Tree canopy

Impervious surfaces

Tree canopy

Percent Difference

Population-based scenarios: Impervious surfaces vs. Tree canopy

Impervious surfaces

Tree canopy

Interactive maps (IRR)

A couple of ideas.

Population-based: Imp vs Tree (side by side)

Population-based: Imp vs Tree (separate layers)

Please navigate to the layer icon under the Zoom icon and select the layer corresponding to each type of intervention.